Artistic Multiprojection Rendering

Artistic Multiprojection Rendering Maneesh Agrawala∗ Denis Zorin† Tamara Munzner∗ ∗Stanford University †New York University Abstract In composing hand-drawn images of 3D scenes, artists often alter the projection for each object in the scene independently, thereby generating multiprojection images. We present a tool for creating such multiprojection images and animations, consisting of two parts: a multiprojection rendering algorithm and an interactive interface for attaching local cameras to the scene geometry. We describe a new set of techniques for resolving visibility between geometry rendered with different local cameras. We also develop several camera constraints that are useful when initially setting local camera parameters and when animating the scene. We demonstrate applications of our methods for generating a variety of artistic effects in still images and in animations. 1 Introduction In computer graphics we typically use a single linear projection – often a perspective projection – to generate a realistic view of a scene. Linear projections achieve this realism at the cost of imposing restrictions on the 2D shape of each object in the image and on the overall composition of the picture. Artists have developed a variety of techniques for composing images of 3D scenes that deviate from the standard perspective projection. One of the most common techniques is to combine multiple projections in a single image. Artists create such multiprojection images for several reasons, including: expressing a mood, feeling or idea; improving the representation or comprehensibility of the scene; and visualizing information about the spatial relationships and structure of the scene. Multiple projections could similarly enhance computer-generated images and animations, but simple and efficient methods for multiprojection rendering have not been available. Today, the most common method for creating a multiprojection image requires a combination of 3D rendering and 2D image compositing. The process is a labor intensive cycle that involves rendering multiple views of the scene, transferring the images into a compositing application and then manually merging them into a single multiprojection image. Although some research systems [12, 11] shortcut this cycle by combining rendering and compositing into a single application, resolving visibility between the images remains a manual process. In this paper we present interactive methods for creating multiprojection images and animations. The main technical contributions of our work are new algorithms designed for: Resolving Visibility: In the multiprojection setting there is no uniquely defined solution to the visibility problem. However, in many cases the user wishes to maintain the visibility ordering of a master camera while using different local cameras to introduce shape distortions to individual objects. Based on this insight, we propose an algorithm that automatically resolves visibility for most practical cases and allows user adjustments when the automatically computed visibility is not satisfactory. Constraining Cameras: We suggest a simple and intuitive set of camera constraints allow- ing the user to choose appropriate projections for a variety of artistic effects. These constraints are particularly effective when initially placing cameras and when animating the scene. Interactive Rendering: We leverage multipass hardware rendering to achieve interactive rendering rates. The user can immediately see how changing the parameters of any camera or moving any object will affect the final image. The remainder of this paper is organized as follows. In section 2, we describe artistic uses of 1 multiple projections. After presenting related work in section 3, we describe our multiprojection rendering algorithm with an emphasis on resolving occlusion in section 4. In section 5, we present a set of camera constraints that provide intuitive controls over the camera parameters. Examples of images generated with our system appear in section 6. Finally, section 7 outlines future directions and conclusions. 2 Artistic Uses of Multiple Projections Using a single projection for an entire scene is restrictive. The ideal projection for one object may not be the best projection for all objects in the scene. Artists have a long history of solving this problem by creating multiprojection images, which generally serve some combination of three functions: • Artistic Expression – Multiple projections help the artist express a mood, feeling or idea. • Representation – Multiple projections improve the representation or comprehensibility of the scene. • Visualization – Multiple projections communicate information about the structure of the objects and spatial relationships in the scene. In this section, we consider several specific examples of each function. In section 6, we present several images and animations that are based on these examples and were created using our multiprojection rendering tools. 2.1 Artistic Expression Viewing Anomalies: In Giorgio de Chirico’s The Mystery and Melancholy of a Street, figure 1(a), the buildings, the van and the ground plane all have different viewpoints. Willats [19] suggests that the melancholy aura is created by the unusual arrangement of the objects which results in an incongruous spatial system. Despite the large disparities between projections, the overall impression of a three-dimensional space remains intact. Cézanne similarly incorporates multiple viewpoints in Still Life with Fruit Basket, figure 1(b). Loran [13] describes how these viewing distortions generate tension between different planes in the image; the distortions flatten some regions of the picture, while enhancing depth in other regions. He explains that the inconsistencies in projection generate an “emotional nonrealistic illusion of space.” Foreground Elements in Animation: In cel-based animations, moving foreground elements sometimes translate across the image with little to no change in parallax (like cutouts or sprites). Equivalently, as the element translates in the scene, its local camera moves with it, thereby maintaining an identical view of the element from frame to frame. Meanwhile, the background is rendered using its own separate projection. Although fixing the view of foreground elements is done primarily to reuse the same drawing from frame to frame, it has become a stylistic convention in hand-drawn animation since the foreground elements appear flatter and more “cartoony.” In contrast to the flattened foreground elements, animators often exaggerate the perspective projection of the background to create a deeper, more dynamic environment. However, such a strong perspective could introduce distortions in the foreground of the image. A fixed view of the foreground objects alleviates this problem. 2.2 Representation Best Views: Certain viewpoints are better than others for comprehending the overall geomet- ric shape of an object. By using multiple viewpoints the artists can present the best views of all objects in the scene. Graham [5] points out that a single viewpoint is often inadequate for large format pictures like murals, frescoes, or billboards. Such images are often placed above standard eye-level and are seen from a much wider range of viewpoints than smaller format pictures. For these reasons, many large format images are created with multiple projections. As 2 A B D C (a) Giorgio de Chirico’s The Mystery and (b) Cezanne’s Still Life with (c) Paolo Uccelo’s Sir John Melancholy of a Street Fruit Basket (Diagram from [13]) Hawkwood Axonometric Oblique (e) Artificial Perspective via Multiple Parallel Projections (Chairs from [2], 53rd Street from [6]) Fig. 1. Multiprojection examples (clockwise labeling) (a) A different projection is used for each major structure. (d) Raphael’s School of Athens Note the difference in vanishing points for the buildings, the van, and the ground plane. An exaggerated perspective projection elongates the white building and the ground plane is placed so that the horizon is high up in the image to create the long receding path. (b) Cezanne incorporates many viewpoints into this still life, of which four are shown in the diagram. Notice how viewpoint A is much higher than the other viewpoints, thus the ginger jar and basket appear to be tipped forward. Differences in viewpoint also cause the table to appear to split under the table cloth. (c) The projection for the base differs from the projection for the horse and rider. In its original setting viewers stood below and to the left of the fresco, so the entire picture was first created for this viewpoint. While the low−set viewpoint was effective for the base, Uccelo found that it exposed too much of the horse’s belly and distorted the rider. He repainted the top of the picture and raised the viewpoint of the horse and rider to eliminate the distortion. (d) The foreground humans would appear distorted if rendered using the projection of the background architecture. Thus, Raphael altered the projections for the humans to give each one a more central projection. Without the correction, the sphere (inset) would appear elliptical rather than circular[10, 21]. (e) Multiple parallel projections can produce an artificial sense of perspective by pointing the receding parallels towards a central vanishing point. Dubery and Willats[2] argue that this technique works better with oblique projections than with axonometric projections. Each building in the illustration of 53rd Street is projected using a different

Load more